Page 526 - Practical Design Ships and Floating Structures

Practical Design of Ships and Other Floating Structures 50 1
You-Sheng Wu, Wei-Cheng Cui and Guo-Jun Zhou (Eds)
Q 2001 Elscvier Science Ltd. A11 rights reserved

EMPIRICAL PREDICTION OF SHIP RESISTANCE AND WETTED
SURFACE AREA USING ARTIFICIAL NEURAL NETWORKS

Kourosh Koushan

Norwegian Marine Technology Research Institute (MARINTEK)
P.O.Box 4 125 Valentinlyst, 7045 Trondheim, Norway

ABSTRACT
New empirical methods are presented for prediction of ship resistance and wetted surface area of ships
based on analysis of database of tests performed in the towing tank of MARJNTEK. Artificial neural
networks method is applied for analysis of the database. The methods are verified using several towing
test results available. These methods show generally reliable simulation of residual resistance and
wetted surface area of the ships.

KEYWORDS

Ship resistance, Residual resistance, Wetted surface, Empirical method, Artificial neural networks,
Database, Simulation

1 INTRODUCTION
There has been a significant development in the field of numerical calculation of ship resistance in the
recent decade. These have led to useful tools for detail analysis of ships. Inevitably require all these
tools a complete physical description of vessel. However at an early design stage require naval
architects often a tool for reliable prediction of ship resistance based on few main parameters, usually
Froude number and some geometric coefficients. Again there have been remarkable efforts to cope
with this situation and several empirical methods are developed and optimised over the years applying
regression analysis. e.g. Holtrop (1984) and Hollenbach (1998).

Experience has shown that all these methods can predict some cases very well whereas in some other
cases predictions might not be as reliable. The method presented in this paper applies artificial neural
networks (ANN) for the analysis of the extensive database of towing test results performed at
MARINTEK in recent two decades to predict residual resistance coefficient. ANN method allows for
non-linear effects and interdependence of input parameters. The database is divided into several ship
categories and a network is designed for each category, allowing a more accurate prediction for each
category. For the first time a reliable empirical method is developed for prediction of resistance of
offshore vessels and car ferries. Objective of the method is to keep the number of input parameters as
low as possible however at the same time deliver a reliable prediction, which can help the designer at